Fabric defect pixel-level classification method based on deep learning

Abstract

The invention provides a fabric defect pixel-level classification method based on deep learning. The fabric defect pixel-level classification method is specifically implemented according to the following steps: step 1, collecting defective fabric images to form a picture set; step 2, establishing a MobileNetV2 network model; step 3, training the pre-training set by using a MobileNetV2 network model; step 4, establishing a Mobile-Unet network model; step 5, training the training set by using a Mobile-Unet network model; and step 6, classifying the input pictures by the trained Mobile-Unet network model, and outputting the classified images. According to the method, pixel-level segmentation can be carried out on defective fabrics, parameters and models in the method are smaller, and the robustness of the algorithm is improved.

Description

technical field [0001] The invention belongs to the technical field of image segmentation, and relates to a pixel-level classification method for fabric defects based on deep learning. Background technique [0002] The competition in the textile industry is becoming increasingly fierce. The last process after the cloth weaving is usually fabric defect detection, and then evaluates the product grade. of great pressure. Aiming at the detection of fabric surface defects, many domestic and foreign scholars have done related research. These detection methods can be divided into three categories: statistical-based methods, frequency-domain-based methods, model-based methods, and learning-based methods. Statistical-based methods rely on the choice of parameters, which are less accurate and less precise. The detection result of the method based on the frequency domain depends on the selection of the filter, and the detection effect on complex textured fabrics is poor. The model-...

AI Technical Summary

Problems solved by technology

Statistical-based methods rely on the choice of parameters, with poor accuracy and low precision

The detection result of the method based on the frequency domain depends on the selection of the filter, and the detection effect on complex textured fabrics is poor

The model-based method has a large amount of calculation and is difficult to meet the real-time requirements

The above methods have opened up a good way for machine vision detection of texture defects, but further research is needed to find a robust method with strong adaptability to variable textures

Images